TL;DR
From 12 November 2026, Meta's pixel starts measuring some conversions on the customer's device instead of on your server or website. The setting is called "Measure conversions on customer devices" and it is already switched on by default for datasets that send events. Expect your reported conversions to move, probably up, and expect the move to land right in the middle of Black Friday prep. Do not panic, do not rebuild your campaigns, and do take a baseline screenshot now so you can tell a measurement change from a real performance change later.
What actually changes
Until now, your pixel and Conversions API sent events from your site or server to Meta, and Meta matched them to ad views and clicks. From 12 November, some of that measurement happens on the customer's own device, using signals Meta can read there. Meta ran an experiment on this between 6 and 21 August 2026, across roughly 117,700 advertisers, and reported about 3% more conversions on average. That is Meta's number from its own test, not ours, and your mileage will vary.
The setting lives in Events Manager, under your dataset's settings. If your dataset already sends events, the switch is on by default and the panel reads "Measurement starting November 12, 2026". You can turn it off. The question is whether you should.
Why your reported numbers will move
Three things happen at once, and they all look like performance changes when they are not.
First, your reported conversion count will probably rise. On-device measurement catches purchases and leads that server-side matching missed, especially on iOS and in browsers where third-party cookies are already dead. More counted conversions means a lower reported cost per result, even if nothing about your ads or your customers changed.
Second, your cost per result will look better overnight, and someone will take credit for it. If your agency, your freelancer, or your own Monday-morning self reports a sudden improvement on 13 November, check the calendar before celebrating. A 3% lift in counted conversions is not a creative breakthrough.
Third, your historical comparisons break. Any year-on-year or month-on-month comparison that crosses 12 November is now comparing two different counting methods. Your November will look better than your October for a reason that has nothing to do with your Black Friday creative.
What to do before 12 November
This is a thirty-minute job, and the deadline is real because the change lands two weeks before Black Friday.
Take a baseline. Screenshot your last 28 days of reported purchases, cost per purchase, and attributed revenue by dataset, today. Save it somewhere you will find it in January. Without a baseline you cannot separate the measurement change from the Black Friday auction, and December is the worst possible month to guess.
Decide per dataset, not per account. You may run several datasets, and the setting is per dataset. A lead-gen dataset and a purchase dataset are different measurement stories. Decide each one on its own merits, and write the decision down.
Leave it on unless you have a reason not to. The default exists because Meta wants more complete measurement, and more complete measurement is usually good for optimisation. The cases for switching it off are specific: strict consent setups, regulated industries, or a legal team that has actually read the terms. "I don't trust Meta" is a feeling, not a measurement policy.
Tell whoever reads your reports. If a client, a boss, or a finance team sees your numbers, warn them now that reported conversions will shift around 12 November. A five-minute message today beats a panicked call on 14 November asking why the numbers jumped.
What not to do
Do not rebuild campaigns around the change. Your targeting, budgets, and creative do not care where a conversion gets counted. Do not raise budgets because reported cost per result fell, and do not kill ads because a dataset you switched off now reports fewer conversions. The ads did not change. The ruler did.
And do not use the change to hide bad news. If your real performance is sliding into Black Friday, a measurement lift will mask it for about a week. That is the most dangerous version of this story: numbers that look fine while the business quietly gets worse.
FAQ
What is "Measure conversions on customer devices"?
A Meta pixel setting that moves some conversion measurement onto the customer's device, so events Meta previously missed through server-side matching get counted. It is on by default for datasets already sending events, and measurement starts 12 November 2026.
Will my reported conversions go up?
Probably, modestly. Meta's own August 2026 experiment across about 117,700 advertisers measured roughly 3% more conversions on average. Your result depends on your traffic mix, your consent setup, and how much signal you were already losing.
Should I turn the setting off?
Only for a concrete reason, such as a consent or compliance requirement. For most advertisers the extra counted conversions improve optimisation and reporting. If you do switch it off, do it per dataset and note the date, so future comparisons make sense.
How do I tell a measurement change from a real performance change?
Take a baseline of the last 28 days before 12 November, then compare the days after against it while watching your own first-party numbers too. If Meta reports more conversions but your till, CRM, or bank account does not, that is measurement, not growth.
Does this affect Black Friday reporting?
Yes, and that is the awkward part. The change lands 15 days before Black Friday, so your Cyber Week numbers will be counted differently from your October numbers. Baseline now, annotate the date in your reports, and judge Cyber Week against your own sales data first.
MetaMaxd is a Meta ads agency. We read the measurement fine print so your November numbers mean something when the bill arrives. If your reports need a second pair of eyes before Black Friday, that is what we do.
